Open AI CEO Sam Altman testified before Congress earlier this year.
BRENDAN SMIALOWSKI/AFP
hide caption
toggle caption
BRENDAN SMIALOWSKI/AFP
This week, some in Congress seemed to discover a new urgency to act on the complicated task of AI regulation after a slew of grim warnings from industry leaders. The only issue: Most of them leave town this week until after the November election.
And they are not just short on time — they’re short on consensus on whether regulation is even a task for Congress to tackle. Take, for example, the leaders of the parties in the House of Representatives.
Speaker of the House Mike Johnson, R-La., said Tuesday that “AI is serious. We do need to make sure that it’s done safely but let’s again let’s apply common sense.” He then called on the leaders of frontier AI labs to self-regulate.
“They don’t need the government to tell them to slow it down. If they want to slow it down, they should,” Johnson said. “But we cannot have a moratorium on the development of AI. Okay? Because then we will lose our edge to China and that has serious national security implications.”
Democratic Minority Leader Hakeem Jeffries of New York has taken a different tack.
“It requires decisive congressional action immediately in a manner consistent with what some of the leading AI voices in the country are now saying needs to happen as it relates to slowing things down to protect the health, the safety, and the well-being of the American people,” Jeffries said Monday.
The issue also cuts across party lines. Steve Bannon, a key adviser to President Trump, and Vermont progressive Sen. Bernie Sanders spoke separately at an event Tuesday in Washington, calling for the technology to be restricted.
Quick action is unlikely, but lawmakers are not starting from scratch
OpenAI and Anthropic, two of the leading American AI companies, are often at odds on what regulation is necessary. But leaders from both companies recently expressed support for independent watchdogs being able to assess the development process.
Though the issue boiled over this week, AI regulation efforts have been simmering on the Hill for years. A bipartisan House task force released a report in 2024. Earlier this year, Representatives Jay Obernolte, R-Calif., and Lori Trahan, D-Mass., released a comprehensive bipartisan proposal.
Those efforts, along with a number of others, were greeted with a shrug by leaders from both parties.
Republicans lack consensus within their on party on the right approach, Democrats appear ready to wait for now, as they anticipate their party will control of one or both chambers of Congress come January.
A fight over who should do the regulating
If and when regulation efforts do, eventually, begin in earnest, one of the key details lawmakers will have to decide is whether states will be allowed to impose their own regulations or whether their statutes will be preempted by federal rules, as desired by the Trump administration.
“The preemption issue is going to be one of the most difficult things that we navigate,” Rep. Kat Cammack, R-Fla., said at an Axios event in March. “Companies are not going to be able to have a framework for 50 different states and really survive.”
While the GOP has typically supported having states lead on most regulation, House Republican leaders, including Rep. Steve Scalise of Louisiana, have expressed support for broad bans on states’ ability to regulate AI. Many Democrats and states’ rights-minded GOP lawmakers oppose them.
“What we’ve seen over and over again is that while Congress has essentially fiddled under the influence of big tech lobbyists, states like Texas and Florida and Utah, red states in particular,” said Daniel Cochrane, “have been leading the charge to hold these companies accountable.”
Cochrane, who spoke to NPR in June, is a senior fellow for the Family First Tech initiative at the Institute for Family Studies, a research and advocacy group that works to strengthen marriage, family life and the welfare of children. He worries that a broad federal ban on state AI regulation would result in inaction.
“Almost everyone in Washington agrees AI is transformative and requires some kind of policy response,” Nicole Alvarez, a senior tech policy analyst at the Center for American Progress, told NPR this summer.
“The real fight,” she said, “is what governance looks like.”
She worries that, even if a federal framework does move forward, states will be limited if they want to place further restrictions in areas like consumer safety and protections.
While Congress sorts out its approach, Adam Kovacevich, the founder and CEO of center-left tech industry policy group Chamber of Progress worries the patchwork landscape of state regulations is imposing a muddled, de facto national standard.
“Unfortunately we’re seeing a creep of blue state laws” on frontier models, fairness and algorithmic discrimination, Kovacevich told NPR in June, “but that’s not the best way to regulate AI.”
He thinks AI should be regulated largely at the federal level but “there seems to be little energy being spent on actually writing those standards before this year’s midterm elections.”
It is tough to see beyond the fight about how to regulate to the specifics what rules Congress might agree on. Protections for children appear to be one possible area of common ground.
“Some uses of AI are so structurally dangerous or misaligned with basic social values that they shouldn’t just be regulated, they should be prohibited outright,” Alvarez said. “For example, banning AI systems from generating child sexual abuse material.”
Cochrane points to a proposal from Sen. Josh Hawley, R-Mo., the GUARD Act, which he says “would impose restrictions on AI companions for children and impose criminal liability on anyone who makes a sexually explicit chatbot.”
Kovacevich wants to see investments in competitiveness.
“We are in an AI race against China and the Chinese government is viewing AI as an existential race to win,” he said. “They’re making all sorts of government investments and policies to win that race. We’re just not treating the race with the same level of seriousness.”
